Repository navigation
Releases
Swarm Workbench 0.8.1 alpha
Swarm Workbench 0.8.1 alpha
Downloadable native desktop workbench for Windows x64, macOS Apple silicon/Intel, and Linux x64. End users need no Python, Node.js or Rust.
The release adds mission briefs, graph linting and failure branches; reviewed task reuse and provenance-aware artifact assembly; repeated ablations, promotion gates and checkpoint comparisons; encrypted workspace sync; verified inference runtime installation with rollback; private vLLM cluster controls; and isolated local MCP/ACP extensions.
Open Control room → Quality, Control room → Managed systems, and each mission’s Artifacts & checkpoints tab. Mission briefs appear in the new-mission composer, and graph diagnostics/outcome controls appear in the architecture editor. The 22 presets and bounded recursive architecture workshop continue to use actual connected models.
See bundled docs/FEATURES_0_8.md and docs/QUALITY_AND_SYSTEMS.md for controls, limits and integration references. Private model sharding requires explicitly approved trusted Linux GPU workers and a compatible model/runtime/network; physical multi-host performance validation is separate from protocol tests. Extension isolation requires an installed digest-pinned Docker image.
Release artifacts carry updater signatures and SHA-256 checksums. Operating system code signatures/notarization depend on separately provisioned certificates. Model weights, optional inference runtimes and container images are separate downloads. Android assets are development APKs; iOS assets target the simulator.
Swarm Workbench 0.7.0 alpha
Downloadable Swarm Workbench 0.7.0.
Installers carry updater signatures. Operating system signatures depend on separately provisioned certificate credentials. Model weights and optional runtimes are separate downloads.
See bundled docs/FEATURES_0_7.md and docs/BUILD_STATUS.md for supported scope and validation.
Swarm Workbench 0.6.5 alpha
Swarm Workbench 0.6.5 expands the downloadable desktop application's architecture library from ten to twenty-two presets and adds a new visual editor and bounded recursive teams.
- New patterns include map/reduce, routing and specialists, test-driven work, evidence auditing, a judged debate, red/blue teams, specialist reviews, diamond comparisons, and specify/build/verify.
- Recursive decomposition, critique/repair, and fork/join execute real child teams. Self-calls survive custom copies; upstream outputs can scope the next task. Depth limits and stopping markers switch to a connected model that solves the base case.
- Editing adds search/categories, a selectable team list, duplicate/remove, undo/redo, automatic layout, explicit connection controls, reference validation and execution previews. Child-team buttons and parent/root navigation make nested work inspectable.
- Shared root budgets, immutable snapshots, permission intersections and checkpoint reuse keep recursive work bounded. Four nested levels, 128 expanded nodes and a configurable 0–32 child-run limit are supported. Preflight includes nested model stages.
Missions, automations, campaigns, comparisons, and evaluations use connected models. If no eligible model is available, the mission reports missing capacity or waits when the owner enabled queuing. The offline demonstration mode remains removed.
Older demonstration records remain available for inspection and export. Resuming, branching, or chatting with those records is disabled. Onboarding now guides owners toward a provider, local model, or linked device. Automated checks use owned provider fixtures outside the production app.
This release retains the twelve command-center features introduced in 0.6:
- Project campaigns with dependencies, shared budgets, acceptance review and restart recovery.
- Routing feedback, architecture preflight with shared quota estimates, and versioned project knowledge.
- Explicit local voice input, reviewed MCP Apps panels and isolated Docker notebooks.
- Optional local LoRA training, evaluation-gated promotion, rollback and measured power scheduling.
- Live model compatibility packs and a separate native Android/iOS companion.
Download the desktop installer or portable package for your platform. Model weights and optional inference, transcription, Docker and training runtimes are separate installations. Notebook code requires an already installed container image. The training backend requires a separate CUDA environment and downloaded local weights.
The Android companion development APK supports sideload testing. The iOS build targets the simulator; physical iOS distribution needs Apple signing and provisioning. These builds do not imply Play/App Store publication. Provider keys remain on the coordinator, and phone access uses scoped, revocable credentials.
Updater signatures and SHA-256 checksums accompany desktop assets. OS signing/notarization depends on separately provisioned publisher credentials. See the architecture guide, the feature guide, companion setup and bundled BUILD_STATUS.md for supported scope and validation.
Swarm Workbench 0.6.4 alpha
Swarm Workbench 0.6.4 removes the offline demonstration mode from the downloadable desktop application. Missions, automations, campaigns, comparisons, and evaluations now use connected models. If no eligible model is available, the mission reports missing capacity or waits when the owner enabled queuing; it never substitutes scripted answers.
Older demonstration records remain available for inspection and export. Resuming, branching, or chatting with those records is disabled. Onboarding now guides owners toward a provider, local model, or linked device. Automated checks use owned provider fixtures outside the production app.
This release retains the twelve command-center features introduced in 0.6:
- Project campaigns with dependencies, shared budgets, acceptance review and restart recovery.
- Routing feedback, architecture preflight with shared quota estimates, and versioned project knowledge.
- Explicit local voice input, reviewed MCP Apps panels and isolated Docker notebooks.
- Optional local LoRA training, evaluation-gated promotion, rollback and measured power scheduling.
- Live model compatibility packs and a separate native Android/iOS companion.
Download the desktop installer or portable package for your platform. Model weights and optional inference, transcription, Docker and training runtimes are separate installations. Notebook code requires an already installed container image. The training backend requires a separate CUDA environment and downloaded local weights.
The Android companion development APK supports sideload testing. The iOS build targets the simulator; physical iOS distribution needs Apple signing and provisioning. These builds do not imply Play/App Store publication. Provider keys remain on the coordinator, and phone access uses scoped, revocable credentials.
Updater signatures and SHA-256 checksums accompany desktop assets. OS signing/notarization depends on separately provisioned publisher credentials. See the feature guide, companion setup and bundled BUILD_STATUS.md for supported scope and validation.
Validation: 141 backend tests and 27 headless UI flows pass on each of Windows, macOS, and Linux. Four desktop updater signatures, all published checksum entries, and the HTTPS update feed verify. Windows installation acceptance runs on a disposable CI runner. Automated model responses use owned protocol fixtures; physical model/device acceptance remains separate. Android is development-signed, and the iOS archive targets the simulator.
Swarm Workbench 0.6.3 alpha
Swarm Workbench 0.6.3 adds all twelve command-center features to the downloadable desktop application:
- Project campaigns with dependencies, shared budgets, acceptance review and restart recovery.
- Routing feedback, architecture preflight with shared quota estimates, and versioned project knowledge.
- Explicit local voice input, reviewed MCP Apps panels and isolated Docker notebooks.
- Optional local LoRA training, evaluation-gated promotion, rollback and measured power scheduling.
- Live model compatibility packs and a separate native Android/iOS companion.
Download the desktop installer or portable package for your platform. Model weights and optional inference, transcription, Docker and training runtimes are separate installations. Notebook code requires an already installed container image. The training backend requires a separate CUDA environment and downloaded local weights.
The Android companion development APK supports sideload testing. The iOS build targets the simulator; physical iOS distribution needs Apple signing and provisioning. These builds do not imply Play/App Store publication. Provider keys remain on the coordinator, and phone access uses scoped, revocable credentials.
Large TLS certificate bundles remain available as one-use pairing JSON when they cannot fit a QR code. Pairing extracts public certificates from combined PEM files and excludes private keys. Maintenance fixes make microphone Stop release pending startup resources immediately and prevent a cancelled startup from replacing a newer recording. They also fix an ambiguous dependency-card selector in the UI tests and give graph fixtures enough time on loaded runners.
Updater signatures and SHA-256 checksums accompany desktop assets. OS signing/notarization depends on separately provisioned publisher credentials. See the feature guide, companion setup and bundled BUILD_STATUS.md for supported scope and validation.
Validation: 137 Python tests and 26 headless UI flows pass on Windows, macOS and Linux. All four desktop package/frozen-engine/signature checks pass, including disposable Windows install/uninstall acceptance. Published payload hashes, all 26 checksum entries, phone build provenance and the HTTPS update feed verify. Android uses development signing; the iOS download is a simulator build. Physical GPU/container/audio/camera/device acceptance is separate from these fixture checks.
Swarm Workbench 0.5.1 alpha
Downloadable Swarm Workbench 0.5.1. Includes guided setup, reusable sub-swarms, bounded channels, team roles, data destination reviews, response reuse, failure simulations, A2A text tasks, experimental Exo placement and signed community packages.
Installers carry updater signatures. Operating system signatures depend on separately provisioned certificate credentials. Model weights and optional runtimes are separate downloads.
See bundled docs/FEATURES_0_5.md and docs/BUILD_STATUS.md for supported scope and validation.